Shenzhen, China – [Date of Publication] – In a significant boost for open-source AI development, Huawei’s Ascend platform has successfully adapted the Step-Video and Step-Audio open-source large models developed by Step-Ahead AI. These models are now available on the Modelers (魔乐) community, a platform jointly built by Tianyi Cloud and Huawei. This integration provides developers and enterprises with seamless access to powerful multimodal AI capabilities, fostering innovation and accelerating the adoption of AI solutions.
The Modelers community announced today the availability of Step-Ahead AI’s self-developed Step-Video video generation and Step-Audio speech models. The adaptation was achieved using Huawei’s Ascend CANN heterogeneous computing architecture and Ascend servers, ensuring optimal performance and efficiency.
This collaboration marks a crucial step in democratizing access to cutting-edge AI technology. Developers and enterprise users can now directly download and experience these models within the Modelers community. This eliminates the complexities of environment setup and allows for rapid verification of AI solutions.
Step-Video-T2V: A Global Leader in Open-Source Video Generation
Step-Video-T2V stands out as the world’s largest open-source video generation model, boasting an impressive 30 billion parameters. It is capable of generating high-quality videos with 204 frames and a resolution of 540P. According to Step-Ahead AI’s evaluation results, Step-Video-T2V significantly surpasses existing best-in-class open-source video models in areas such as instruction following, motion smoothness, physical plausibility, and aesthetic appeal.
Step-Audio: Pioneering Open-Source Speech Interaction
Step-Audio is the industry’s first product-level open-source large speech interaction model. It can generate speech with a wide range of emotions, dialects, languages, singing styles, and personalized characteristics. This opens up exciting possibilities for creating more natural and engaging voice-based applications.
Expanding the Open-Source Ecosystem
Step-Ahead AI revealed at a recent ecosystem open day that it plans to open-source a text-to-video model in March, further expanding the availability of its advanced AI capabilities.
The Modelers Community: A Hub for AI Innovation
The Modelers community, a joint effort between Tianyi Cloud and Huawei, serves as a one-stop platform for AI developers and technology enthusiasts. It provides comprehensive services including model and dataset hosting, development toolchains, and application showcasing. To facilitate model and application demonstration, the community offers free computing power, enabling users to experience model inference online without the need for complex environment setup.
This collaboration between Huawei, Step-Ahead AI, and the Modelers community represents a significant advancement in the open-source AI landscape. By providing access to powerful models and readily available computing resources, it empowers developers and businesses to explore new possibilities and accelerate the development of innovative AI-driven solutions. The integration of Step-Video and Step-Audio on the Modelers community is poised to fuel further advancements in video generation, speech interaction, and the broader AI ecosystem.
References:
- 机器之心 (Machine Heart). (2024). 华为昇腾适配阶跃星辰Step-Video和Step-Audio开源大模型，上线魔乐社区. [Huawei Ascend Adapts Step-Video and Step-Audio Open-Source Large Models, Launching on the Modelers Community]. Retrieved from [Insert Original Article URL Here]
[Note: Please replace [Date of Publication] and [Insert Original Article URL Here] with the actual date and URL of the original article.]
Views: 0